Attendees
Anjum, Kashif, Les, Joun. Raja and Umar were unable to be present.
General
We had a meeting with UNIMAS and UM on 11/15/2012.
UNIMAS is monitoring other Malaysian and Indonesian hosts outside the regular Beacons. There was a problem with nort meausing the individual pins. this was caused by a change in the format of the ping response. This has been fixed in the master copy at SLAC (but not in the tarball) and UNIMAS.
I have fixed the web site providing script documentation. It is at http://www-iepm.slac.stanford.edu/pinger/scripttable.html
. The problem was a failing web service.
I have extended ping_data_plot.pl to enable showing more days. I also fixed connectivity.pl to showw all todays data so far.
From Arshad: Rector has agreed to support one person visit to SLAC. Please select the best candidate. Arshad is away in either the UK or Korea. When he returns Anjum will get more information. It could be 6 months or a year, and will probably be Kashif. This should be settled in by Nov 21, 2012.
Pakistani Hosts
AIR has an IP address problem.
Pakistani monitors: good=32, no_data=8, partial_data=8, problems=16, total=48,
PERN monitors: good=8, no_data=2, partial_data=3, total=13, problems=5,
Pingerpwr had an IP address issue (they need to change their address). This has been fixed. PIEAS and IIU have new installations in progress. They are both down. PIEAS has a security concern. IIU were going to do their own install, then requested Joun to do the install but they have not provided access to Joun, The SAU strike is over but they are short of IP addresses. NUST will work with HEC to get an IP address..
The problem of traffic between FSBD and Quetta going via Europe due to not having a Pakistan IXP has been by passed by keeping all traffic both ways on WorldCall.
Since there is no IXP in Pakistan so the peering is done abroad. Anwar Amjad (of HEC) would like NUST to put together a workshop/meeting to look at providing a neutral point for an IXP for PTCL, SuperNet and WorldCom. If the IXP is at HEC then it will need to upgrade all its links to PERN-II so it can carry the extra traffic. If this is funded it could take 4-6 months.
Setting up an IXP for Malaysia may also be an interesting topic for the workshop in December,
Non Pakistani Hosts
I am actively working with Algeria. I have asked for an update from UCSD. UERJ/Brazil has a routing problem.
TULIP
Raja has been busy defending his proposal on (Nov 5th), so there is no progress on TULIP. A suggestion taht came out of the defense was to use TBG constraints. However this is very time consuming. So how to use it optimally will take some study.
Raja understands the CGI code in Matlab. He is working on the MatLab version to add a lookup table to allow the choice of Alpha to be based on region and min RTT. When he has this working he will work on porting from MatLab to Mathematica.
It may also make sense to look at using a modified outer circle. I.e. instead of distance (radius in km)= 100km/ms*min_RTT(ms), use distances(radius in km)=alpha_max*100km/mx*min_RTT(ms) where alpha_max is the maximum alpha for landmarks to target in a given region.
Statistical analysis of Alpha
Raja has written a report on the alpha findings from his analysis. I made some suggestions but have not heard anything.
At the start you say identify Africa as one of the regions. However there is no data on Africa. You might add something on why you chose to focus on Europe, N. America and Pakistan. My understanding is that for other regions the main reason was there were insufficient landmarks to be able to make reasonable estimates (i.e TULIP was very inaccurate). Were there other reasons. Do you have any data to justify this conclusion (e.g. the inaccuracy of TULIP).
Your conclusion indicates we can use a range of alphas that depends on the min-RTT and region. This is instead of using an upper range of alpha of 1 as is done in the original CBG. You provide a table of min, max alpha for the 3 regions. Can you extend this to recommended ranges of alpha not just by region but also for min-RTT, e.g. provide a table or the min alpha and max alpha as function of min-RTT and region. How do you get these min and max values. Are they the max and min possible or say the 10% and 90%
In one of your analyses you showed a more detailed graph for Pakistan (attached). You might want to show this. Does it make anything more apparent?
Future meeting time - Les
Next meeting Wednesday Dec 5th 8:00pm Pacific Daylight Time, Thursday Dec 6th 9:00pm Pakistan time.